3

I try to call an R object from Python inside a Quarto document:

---
title: "pandas"
format: html
jupyter: python3
---

```{r}
data("penguins", package = "palmerpenguins")
```

```{python}
penguins=r.penguins
penguins
```

When I execute the chunks one by one in RStudio, everything is okay:

> data("penguins", package = "palmerpenguins")
> reticulate::repl_python() # automatically executed by RStudio
Python 3.10.4 (/Users/.../3.10.4/bin/python3.10)
Reticulate 1.24 REPL -- A Python interpreter in R.
Enter 'exit' or 'quit' to exit the REPL and return to R.
>>> penguins=r.penguins
>>> penguins
       species     island  bill_length_mm  ...  body_mass_g     sex  year
0       Adelie  Torgersen            39.1  ...         3750    male  2007
1       Adelie  Torgersen            39.5  ...         3800  female  2007
...

However, when I try to render this document, it errors this:

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Input In [1], in <cell line: 2>()
      1 # Python chunk
----> 2 penguins=r.penguins
      3 penguins

NameError: name 'r' is not defined

According to RMarkdown documentation, nothing else is required (so no e.g. rpy2).

I try to add library(reticulate) or reticulate::repl_python() in the R chunk but it doesn't solve the issue.

Note: I'm aware of an old unanswered similar question for RMarkdown.

Thanks!

abichat
  • 2,317
  • 2
  • 21
  • 39
  • 1
    I didn't use Quarto, but ran your code (saved as `test-reticulate.Rmd` after replacing `penguins` with `mtcars`) and didn't get any error messages by rendering. I used the following code to render the document: `library(rmarkdown); render("test-reticulate.Rmd")`. My platform from `sessionInfo()` and `sys.version`, respectively. R: `R version 4.1.3 (2022-03-10), Platform: x86_64-conda-linux-gnu (64-bit), Running under: Amazon Linux AMI 2018.03`; Python: `'3.10.4 | packaged by conda-forge | (main, Mar 24 2022, 17:51:59) [GCC 10.3.0]'`. R package versions: `reticulate_1.24 rmarkdown_2.13`. – mastropi May 02 '22 at 06:40
  • Thanks for your attempt, I found a 'trick' that made it works – abichat May 04 '22 at 12:58

2 Answers2

5

Quarto have two engines for render, knitr and jupyter.A related document is here.

If we use:

---
title: "pandas"
format: html
---

```{r}
data("penguins", package = "palmerpenguins")
```

```{python}
penguins=r.penguins
penguins
```

The engine will be knitr. And while render, knitr will use reticulate(R Interface to Python) to run python code chunk. In this process, knitr will do some magic things to convert the form like r.penguins to the form of reticulate. So the document will be successfully rendered. In other words, knitr made some adaptations to let us could easily run python code chunk with reticulate, and if we don't use knitr engine, we can't use the form like r.penguins.

Quarto use r run all code chunks (auto use r package reticulate to run python chunks) when it uses knitr engine.

And Quarto use python to run all code chunks when it uses jupyter (jupyter: python3) engine. If we want run r code, we must use module (such as rpy2) in python chunk (not r chunk, codes in r chunk will not be run).

We can also use r to run all code chunks by setting jupyter: ir (if we have installed IRkernel).But codes in python chunk will not be run. We must use package (such as reticulate) in r chunk to run python code.

This is my understanding. My English is not good, so if some sentence made you puzzled, we could discuss further.

Gleko Ma
  • 78
  • 6
0

I have no explanation, but removing the line jupyter: python3 in the YAML solved my issue.

abichat
  • 2,317
  • 2
  • 21
  • 39